저작권법에따른이용자의권리는위의내용에의하여영향을받지않습니다.

Transcription

1 저작자표시 - 비영리 - 변경금지 2.0 대한민국 이용자는아래의조건을따르는경우에한하여자유롭게 이저작물을복제, 배포, 전송, 전시, 공연및방송할수있습니다. 다음과같은조건을따라야합니다 : 저작자표시. 귀하는원저작자를표시하여야합니다. 비영리. 귀하는이저작물을영리목적으로이용할수없습니다. 변경금지. 귀하는이저작물을개작, 변형또는가공할수없습니다. 귀하는, 이저작물의재이용이나배포의경우, 이저작물에적용된이용허락조건을명확하게나타내어야합니다. 저작권자로부터별도의허가를받으면이러한조건들은적용되지않습니다. 저작권법에따른이용자의권리는위의내용에의하여영향을받지않습니다. 이것은이용허락규약 (Legal Code) 을이해하기쉽게요약한것입니다. Disclaimer

2 보건학석사학위논문 Comparison between Time Integrated Sampling and Direct Reading Sampling of Nanoparticulate at the Workplace 작업장의나노입자상물질에대한 시간누적시료채취와실시간시료채취방법비교 2015 년 2 월 서울대학교보건대학원 환경보건학과산업보건전공 김선주

3 Comparison between Time Integrated Sampling and Direct Reading Sampling of Nanoparticulate at the Workplace 작업장의나노입자상물질에대한 시간누적시료채취와실시간시료채취방법비교 지도교수윤충식 이논문을보건학석사학위논문으로제출함 2014 년 10 월 서울대학교보건대학원 환경보건학과산업보건전공 김선주 김선주의보건학석사학위논문을인준함 2014 년 12 월 위원장이승묵 ( 인 ) 부위원장최경호 ( 인 ) 위원윤충식 ( 인 )

4 Abstract Comparison between Time Integrated Sampling and Direct Reading Sampling of Nanoparticulate at the Workplace Sunju Kim Department of Environmental Health Graduate School of Public Health Seoul National University, Korea Advisor Chungsik Yoon, Ph.D., CIH Objective Nanoparticles are generated by engineered and unintended in a variety of workplaces and process. Nanoparticles generation in unintended nanoparticle emitting workplace may originate from hot process, welding process. Number and surface area concentration of nanoparticles is generally assessed by a variety of direct reading instruments unlike traditional method. The purposes of this study were to compare time integrated sampling and direct reading instrument sampling method at nanoparticles generation workplaces, and to compare statistically full time sampling and time interval sampling categorized by time integrated sampling method. As characteristics of a variety of workplaces measured in this study were investigated, this study suggests appropriate sampling method by workplace. Methods Sampling was used two methods; the way to use a filter which sampled the air of workplace as time integrated sapling method and the way to sample by - i -

5 direct reading instrument such as SMPS, DustTrak, AeroTrak. Each filter was measured the direct reading instrument simultaneously, and for full time sampling compared with time interval, time interval sampling is measured for working time while replacing a filter by predetermined time interval. Analysis is performed for mass, metals, and TEM. Statistical analysis is performed to compare associations between metrics. Results Concentrations measured at unintended nanoparticle emitting workplaces were higher than those at engineered nanoparticle manufacturing workplaces. CV of concentration is larger as shorter time interval. Full time samples were significantly higher coefficient than time interval samples in spearman s rank test. Conclusions As PM 2.5 concentration out of total mass concentration measured at welding workplaces was above 90%, mass concentration is recommended by gravimetric method. Although concentration measured at engineered nanoparticle manufacturing workplaces was relatively low, nanoparticles generation ratio is high. Therefore, the best way of sampling methods are recommended by using together direct reading instrument and gravimetric sampling method. Keyword: Engineered nanoparticles, Unintended nanoparticles, Direct reading instrument, Time Integrated Sampling, Sampling method Student number: ii -

6 Contents Abstract i Contents iii List of Tables iv List of Figures v 1 Introduction 1 2 Methods and materials Definition of expression Study design Sampling methods Time integrated sampling Direct reading sampling method Analysis methods Statistical analysis 16 3 Results Characteristics of workplaces Comparison of time integrated sampling and direct reading sampling method Comparison of full time and time interval sampling methods Comparison of the full shift and time-interval sampling methods 28 4 Discussions 33 5 Conclusions 38 6 References 39 국문초록 42 - iii -

7 List of Tables Table 1. Summary of workplace characteristics 8 Table 2. Characteristics of direct-reading instruments 14 Table 3. Offline and online concentrations 18 Table 4. Coefficient of variation (CV) based on the results from timeinterval gravimetric sampling and DustTrak direct-reading sampling from one day 27 Table 5. Spearman correlation coefficient result 30 Table 6. Characteristics of online and offline monitoring results according to workplace type 37 - iv -

8 List of Figures Figure 1. Diagram of sampling strategy in this study. 6 Figure 2. The layout of workplaces. 10 Figure 3. Size distributions from SMPS data from each workplace. 21 Figure 4. TEM images from monitored workplaces. 23 Figure 5. Time-integrated sampling results and DustTrak direct-reading sampling results. 26 Figure 6. Regression analysis of full shift and time-interval gravimetric sampling data compared to direct-reading sampling data v -

9 1 Introduction Aerosol sampling methods have long been in use, from the 19 th century, when they were initially developed from sugar tubes, to the invention of time-weighted average (TWA) sampling methods in the 1970s, which are still used today (Spurny, 1998). However, although sampling methods are able to collect smaller-sized particles as newer technology is developed, time-integrated sampling (i.e. gravimetric sampling) is still the reference method used by the National Institute for Occupational Safety and Health (NIOSH), the United States Environmental Protection Agency (EPA), and the Korea Occupational Safety and Health Agency (KOSHA). With recent developments in nanotechnology, it is possible not only to generate engineered nanoparticles (particles <100 nm in diameter) in a variety of workplaces and processes but also to measure nano-sized particles. However, many toxicity studies have shown that the toxicity of nanoparticles, also known as ultrafine particles (UFPs), is greater than that of the same mass of larger particles of similar chemical composition. Many studies have also shown that exposure to nanoparticles causes respiratory and cardiovascular diseases (NIOSH, 2009). Workplaces with nanoparticle exposure are categorized as either unintended nanoparticle emitting workplaces or engineered nanoparticle manufacturing workplaces. Nanoparticle generation in unintended nanoparticle emitting workplaces may originate from hot processes, such as welding (Zimmer et al., 2002), and grinding (Maynard et al., 2002). In previous studies, high particulate - 1 -

10 concentrations have been reported in workplaces where activities such as hightemperature and welding processes take place (Zimmer et al., 2002; Maynard et al., 2002; Elihn et al., 2009). However, sampling of nanoparticles is limited to assessment by gravimetric sampling methods, which is the traditional method (Heitbrink et al., 2009). A reference method for nanoparticle exposure assessment has not yet been determined; however, assessment methods generally use three metrics: number, surface area, and mass concentration. Instead of using the traditional method, number and surface area concentration are normally assessed by a variety of directreading instruments such as a condensation particle counter (CPC), scanning mobility particle sizer (SMPS), or DustTrak aerosol monitor. Mass concentrations of nanoparticles are generally underestimated compared to larger particles (Heitbrink et al., 2009). Although gravimetric sampling is the reference method, it cannot identify variations in concentration because particles are collected over a long period of time (Morawska et al., 2003). The TWA method over an 8-hour period as a timeintegrated sampling method may underestimate particulate exposure in the workplace because the variation in exposure during the work day is large, ranging from background concentrations, such as during break time, to outdoor concentrations and working concentrations. However, unlike direct-reading instruments, with traditional sampling it is possible to identify the chemical composition, morphology, and size of particles. One drawback of direct-reading instruments is that accuracy and reliability are still limited. Few studies have - 2 -

11 compared measurements from time-integrated sampling methods, such as fine particulate matter (PM 2.5 ) concentrations, to those from direct-reading instruments, such as a mass-based DustTrak (Kim et al., 2004; Zhu et al., 2011). The purposes of this study were to compare time-integrated sampling methods and direct-reading instrument sampling methods at nanoparticle generating workplaces, and to statistically compare direct-reading instrument sampling methods and time-integrated sampling methods. Also, as we investigated the characteristics of a variety of workplaces in this study, we suggest appropriate sampling methods according to workplace type

12 2 Methods and materials 2.1 Definition of expression Offline monitoring: gravimetric sampling Online monitoring: direct reading sampling CV (Coefficient of Variation): ratio of the standard deviation to the mean 1 min (1 min for DustTrak; Table 4): 1 min is collected by DustTrak data in the same time periods with gravimetric sample (DustTrak logging time is 1 min). TIS time (Time Interval Sampling time for DustTrak; Table 4): TIS is selected by DustTrak data that calculated by geometric mean respectively in the same time periods with gravimetric sample time. TIS sample numbers are the same with gravimetric those

13 2.2 Study design This study was conducted at two engineered nanoparticle manufacturing workplaces (workplaces A and B) and at two unintended nanoparticle emitting workplaces (workplaces C and D) where welding processes took place. Measurements were performed for 2 days at each workplace. Sampling was conducted during two full working shifts and once overnight at all workplaces, except for workplace B, which was sampled for one full working shift and once overnight. The complete sampling strategy design is shown in Figure

14 Figure 1. Diagram of sampling strategy in this study

15 Workers in workplaces A and B, engineered nanoparticle manufacturing workplaces, were exposed to particles during processes such as bagging and collecting. In workplace A, amorphous silica of 7-40 nm average diameter was manufactured, and each process was an automatic system, with the exception of bagging, which was measured in this study. At workplace B, metallic nanopowders such as aluminum, iron, copper, nickel, and silver were manufactured. Nanopowders were produced by the pulsed wire evaporation method, and the main products were copper-nickel and nickel nanopowder. In workplaces C and D, unintended nanoparticle emitting workplaces with high temperatures, particles were generated from fumes generated during welding and melting processes. An automobile engine part was manufactured in workplace C, and sampling was conducted at the melting process on the first day and at the welding process on the second day. Body frames for heavy equipment, such as forklifts, were manufactured in workplace D, which was located at an industrial complex. Sampling was measured at the welding process

16 Table 1. Summary of workplace characteristics Type of nanoparticle Workplace Nanoparticle generation process Daily production rate No. of workers exposed to nanoparticles/ Total no. of workers Work type Working time per day (hours) Temperature, humidity Engineered nanoparticles (ENP) Unintended nanoparticles A bagging 25 tons 4/1500 semi-automatic 8 (shift work) 31 C, 57% B sieving, collecting 700 g 6/14 semi-automatic 8 26 C, 53% C melting, welding - 63/178 manual 8 (shift work) 35 C, 40% D welding - 30/58 manual 8 (shift work) 41 C, 32% - 8 -

17 (a) (b) - 9 -

18 (c) (d) Figure 2. The layout of workplaces (a) Workplace A, (b) Workplace B, (c) Workplace C, (d) Workplace D

19 2.3 Sampling methods Sampling was measured using two methods. The first method was timeintegrated sampling using filters categorized for mass analysis, metal analysis, and electron microscope analysis. The second method used direct-reading instruments categorized by measurement metrics Time integrated sampling The time-integrated sampling used filters to measure mass concentrations and identify metal contents for quantitative analysis. An electron microscope was used to determine the size, morphology, and aggregation of particles for qualitative analysis. Polyvinyl chloride (PVC Filter, 37 mm, 5 μm, Millipore, Germany) was used for gravimetric analysis according to NIOSH method no. 0500, and mixed cellulose ester membrane filters (MCE Filter, 37 mm, 0.45 μm, Millipore, Germany) were used for metal analysis according to NIOSH method no A transmission electron microscopy (TEM) grid (Q225-CR1, 200 mesh copper, EMS, USA) was used to analyze particle sizes and morphologies with an electron microscope. Each filter attached to the three-piece cassette sampled using a 2-liter per minute (LPM) pump (Gilian Inc., USA). Measurements with direct-reading instruments were obtained simultaneously with the time-integrated sampling. To compare full shift sampling to time-interval sampling, time-interval sampling was conducted for a working shift while replacing the filter at set time intervals

20 The following equation was used to estimate the minimum sampling volume for workplace measurements and evaluations: ( ) = ( μg ) ( / m3 ) (eq. 1), where Ve is the estimated minimum sampling volume, LOQ is the limit of quantification of the measuring device, which is generally threefold the limit of detection (LOD), and Q is the estimated concentration in the workplace

21 2.3.2 Direct reading sampling method Table 2 shows characteristics of the direct-reading instruments. A Nanoscan SMPS (3910, TSI, USA) was used to measure particle number concentrations, a DustTrak aerosol monitor (8533 TSI, USA) was used to measure mass concentrations, and an AeroTrak nanoparticle aerosol monitor (9000, TSI, USA) was used to measure surface concentrations. To confirm the characteristics of the particulates generated during the work, researchers directly recorded in time activity diary (TAD)

22 Table 2. Characteristics of direct-reading instruments Metric Device Remarks Mass DustTrak 8533 (TSI Inc.) ㆍ Particle size range: 0.1 μm to 15 μm ㆍ PM 1.0, PM 2.5, PM 10, Respirable mass Number Nanoscan 3910 (TSI Inc.) ㆍ Particle size range: 10 nm to 420 nm ㆍ 13 channels ㆍ CPC + DMA Surface Area Nanoparticle Aerosol monitor 9000 (TSI Inc.) ㆍ Particle size range: 10 nm to 1,000 nm ㆍ TPM: 1 μm 2 /cc to 2,500 μm 2 /cc RPM: 1 μm 2 /cc to 10,000 μm 2 /cc ㆍ Resolution: 0.1 μm 2 /cc

23 2.4 Analysis methods The PVC filters were stored in a constant temperature/humidity chamber for at least 24 hours (20 C±5, 55%±5), and then weighed using an analytical balance (Ultra Microbalance XP2U, METTLER TOLEDO, USA) to calculate the mass concentration. MCE filters were placed into graphite sample decomposition blocks (i.e, vessel; ECOPRE, ODLAB, Korea), 3 ml of ashing acid (nitric acid and perchloric acid 4: 1 mixtures) was added to the vessel, and the temperature was maintained at 130 C for 30 min. Pretreated samples with a total capacity of 40 ml were analyzed by inductively coupled plasma atomic emission spectrometry (ICP-AES; Model Optima 3100 DV, Perkin-Elmer, USA). During analysis, the appropriate wavelength for each metal was selected according to NIOSH method no The target material in workplace A was silica, but metal content was used as a substitute for mass concentration due to the analysis limitations of ICP. The target materials in workplace B were copper and nickel, and the target materials in workplaces C and D were iron, manganese, chromium, copper, nickel, and zinc. The total metal content was calculated as the sum of the analytical results of each target material. Electron microscopy was used to identify the morphologies and sizes of particles. The TEM-grid samples from the workplaces were analyzed using high resolution (HR)-TEM (High Resolution - Transmission Electron Microscope, JEM-3010, Japan). The elemental analysis of the particles was performed using energy dispersive spectroscopy (EDS; Energy Dispersive Spectrometer, Oxford, UK)

24 2.5 Statistical analysis To compare the time-integrated and direct-reading sampling methods, all directreading instruments were placed next to the time-integrated sampling equipment and measurements were obtained simultaneously. A total of 42 paired datasets were collected. Eight pairs were excluded due to incomplete direct-reading or timeintegrated measurements. Grubbs test was used to test for outliers; one sample was excluded as an outlier, leaving a total of 33 samples for further analysis. The directreading instrument dataset was compared to the time-integrated sampling dataset using the Kolmogorov-Smirnov test. All data were positively skewed, and therefore all of the direct-reading data were adapted to a log-normal distribution. The Wilcoxon signed-rank test was used to compare the full shift and timeinterval time-integrated sampling methods to the direct-reading sampling method. In addition, Spearman rank correlation coefficient was used to identify associations between the time-integrated and direct-reading sampling methods. Linear regression modeling was performed to determine the coefficient of determination. All statistical analyses were performed using SPSS version 20.0 (IBM Inc., USA)

25 3 Results 3.1 Characteristics of workplaces A summary of the direct-reading sampling and time-integrated sampling is provided in Table 3. The mass concentrations during offline monitoring were generally higher in the unintended nanoparticle emitting workplaces than in the engineered nanoparticle manufacturing workplaces. Some metal content samples from workplace B were below the limit of detection (LOD) due to the small output. The total metal contents at workplaces C and D were higher than at workplaces A and B. In addition, metal contents measured at the welding process at workplace D were higher than at the welding process at workplace C. For the online monitoring results, the DustTrak PM 2.5 /total particulate matter (TPM) ratios were >0.9 in workplaces C and D during the welding process, while the ratio was <0.9 during the melting process at workplace C. The highest SMPS measurement of the nanoparticle ratio (> 0.8) was observed at workplace C. The highest surface area concentration was observed at workplace D. As demonstrated by the SMPS measurements of nanoparticle ratios, the engineered nanoparticle manufacturing workplaces generated high nanoparticle number concentrations and high mass concentrations for particles > PM

26 Table 3. Offline and online concentrations Offline Monitoring Online Monitoring Measurement Workplace Mass Metal DustTrak (mg/m 3 ) SMPS (#/cm 3 ) AeroTrak types concentration content* PM (mg/m 3 ) (mg/m 3 ( μm ) PM 2 /cm Total 2.5 <100 > 100 <100 nm Total ) /Total nm nm /Total A Full ,450 6,016 21, Full ,832 3,309 8, Full ,652 1,733 7, interval ,201 4,110 12, interval ,612 9,659 44, B Full ,671 1,727 11, Full < LOD ,253 1,809 7, interval ,193 1,797 11, interval < LOD ,749 1,517 6, C Full ,539 12, , Full ,172 12, , Full ,519 18, , interval ,699 20, , interval ,949 8,674 98, interval ,226 14, , interval ,190 17, , interval ,655 14, , interval ,547 22, , interval ,215 22, , interval ,733 15, ,

27 Offline Monitoring Online Monitoring Measurement Workplace Mass Metal DustTrak (mg/m 3 ) SMPS (#/cm 3 ) AeroTrak types concentration content* PM (mg/m 3 ) (mg/m 3 ) PM ( μm 2 /cm Total 2.5 <100 > 100 <100 nm Total ) /Total nm nm /Total D Full ,231 47, , Full ,859 18,932 63, Full ,553 40, , interval ,446 80, , ,374 interval ,707 21,064 62, interval ,935 92, , ,539 interval ,916 70, , ,084 interval ,439 59, , ,122 interval ,006 47, , interval ,222 9,370 29, interval ,020 56, , ,093 interval ,286 51, , interval ,655 28,216 76, *Sum of target materials: workplace A = Si; workplace B = Cu, Ni; workplaces C and D = Fe, Mn, Cr, Cu, Ni, Zn; LOD (µg/sample): (Cu), (Cr), (Ni), (Mn), (Fe), (Zn)

28 Figure 3 shows the size distributions of SMPS particle measurements for each workplace. The nanoparticle generation ratio was the highest at workplace C. The graph indicates bimodal distributions of nanoparticles at the unintended nanoparticle emitting workplaces and normal distributions at the engineered nanoparticle manufacturing workplaces

29 Figure 3. Size distributions from SMPS data from each workplace

30 Figure 4 shows a TEM image for particle morphology, size, and aggregation analysis by electron microscopy. Small sized particles formed aggregates at the engineered nanoparticle manufacturing workplaces. There were both nanoparticles and larger particles observed at the melting process in workplace C, while almost all particles were nanoparticles at workplace D

31 x 10,000 x 50,000 x 10,000 x 50,000 (a) (b) x 10,000 x 50,000 x 10,000 x 50,000 (c) (d) Figure 4. TEM images from monitored workplaces (a) workplace A, (b) workplace B, (c) workplace C, (d) workplace D

32 3.2 Comparison of time integrated sampling and direct reading sampling method Figure 5 shows the results of gravimetric sampling and direct-reading sampling with the DustTrak for the entire sampling period. Table 4 shows coefficient of variation (CV) using the results of the time-interval gravimetric sampling and the DustTrak sampling measurements from one day. Time interval sampling (TIS) was measured only two samples for one day in workplace A and B, and for two days in workplace C and D. 1 min and TIS time are collected by direct reading sample. The reason for classifying direct-reading instrument data in two ways was to compare variation between 1 min and TIS time intervals. First, gravimetric sampling data was compared to 1 min interval direct-reading data. As shown in Figure 5, CV of gravimetric sampling is observed to lower value compared with 1 min interval data, and the CV of gravimetric sampling compared with 1 min interval in workplace B is higher. CV values of 1 min and TIS time measured by direct reading instrument in workplace A are higher than that of gravimetric sampling. All of direct reading instrument data on the first day in workplace C shows the highest variation compared with the others workplaces because workplace C shows the highest variation on first day measured at melting furnace working for 24 hours (> 1.0). CV of gravimetric sampling is higher than that of 1 min and TIS time measured by DustTrak on the first and second day in workplace D

33 Secondly, as compares with 1 min and TIS time, all of 1 min data shows the higher variation than TIS time. There is a significantly difference of variation between 1.3 times (DustTrak PM 2.5 ) in workplace D on second day and over 16 times (SMPS > 100 nm) in workplace C on first day

34 (a) (b) (c) (d) Figure 5. Time-integrated sampling results and DustTrak direct-reading sampling results (a) workplace A, (b) workplace B, (c) workplace C, (d) workplace D

35 Table 4. Coefficient of variation (CV) based on the results from time-interval gravimetric sampling and DustTrak direct-reading sampling from one day Workplace Gravimetric Sampling DustTrak PM 2.5 DustTrak Total SMPS <100 nm SMPS Total AeroTrak 1 min * TIS time ** 1 min TIS time 1 min TIS time 1 min TIS time 1 min TIS time A B C Day Day D Day Day * 1 min (1 min for DustTrak) is collected by DustTrak data in the same time periods with gravimetric sample (DustTrak logging time is 1 min). ** TIS time (Time Interval Sample time for DustTrak; Table 4) is selected by DustTrak data that calculated by geometric mean respectively in the same time periods with gravimetric sample time. TIS sample numbers is the same with gravimetric those

36 3.3 Comparison of full time and time interval sampling methods Comparison of the full shift and timeinterval sampling methods Sample volumes were calculated using equation 1, which is described in the methods section. The LOD of the balance used for measurements in this study was 6.9 µg, and the lowest concentration during the full shift was mg/m 3. Therefore, the estimated minimum sampling volume was 223 L (the sampling flow rate was 2 L/min), and the minimum sampling time was 1 hour, 55 min. In this study measurements were sampled for a minimum 3 hours at workplaces A and B. At workplaces C and D, the highest concentration during the full shift was mg/m 3, and the minimum sampling time was 4 min. At these workplaces, timeinterval sampling time was appropriate. Gravimetric sampling data were categorized as either full shift samples (11 samples), or time-interval samples (22 samples). The 33 samples of direct-reading instrument data, including gravimetric mass concentrations, AeroTrak surface concentrations, DustTrak PM 2.5 /TPM ratios, and SMPS nanoparticle/tpm ratios, were not normally distributed. Therefore, nonparametric methods were used to compare gravimetric sampling measurements and direct-reading instrument sampling measurements. The Wilcoxon signed-rank test was used to compare gravimetric sampling measurements and direct-reading sampling measurements. The Wilcoxon signed

37 rank test indicated that there were no differences in PM 2.5 and TPM concentrations obtained from gravimetric sampling and DustTrak sampling, but differences were detected between gravimetric and AeroTrak (p <0.001) surface concentrations, and between gravimetric and SMPS TPM concentrations (p <0.001). Spearman s rank correlation was used to determine the strength of the association between gravimetric sampling concentrations, categorized as total gravimetric sampling data, full shift samples, and time-interval samples, and directreading instrument concentrations. Table 5 shows the Spearman s rank correlation coefficient results. The Spearman s rank correlation coefficients for all data were found to be statistically significant (p <0.01). The Spearman s rank correlation coefficient for full shift gravimetric sampling concentrations and SMPS data was >0.9, indicating a strong positive association between the two measurements

38 Table 5. Spearman correlation coefficient result Total Full time Time interval AeroTrak 0.620* 0.782* 0.634* DustTrak PM * 0.736* 0.628* DustTrak Total 0.642* 0.736* 0.675* SMPS <100 nm 0.769* 0.927* 0.652* SMPS Total 0.761* 0.918* 0.679* * p <

39 Linear regression analysis was used to determine the coefficient of determination between the full shift and time-interval gravimetric sampling methods. Figure 6 shows regression model graph, including the coefficients of determination. The time-interval samples had higher coefficients than the full shift samples for all metrics

40 (a) (b) Figure 6. Regression analysis of full shift and time-interval gravimetric sampling data compared to direct-reading sampling data (a) Gravimetric data compared to DustTrak data, (b) Gravimetric data compared to SMPS data, (c) Gravimetric data compared to AeroTrak data. (c)

41 4 Discussions In this study, time-integrated sampling (gravimetric sampling) and direct-reading sampling methods were compared at nanoparticle generating workplaces. Sampling at engineered nanoparticle manufacturing workplaces was conducted at the bagging process because the processes for producing nanoparticles were conducted at contained facilities. Although the PM 2.5 /TPM ratios ( ) at these workplaces were lower than that at other workplaces, the nanoparticle (i.e., UFP) ratios were not low. These results suggest that particles were already aggregated at the bagging process. The TEM images showed that nanoparticles were aggregated, and nanoparticles have a natural tendency to agglomerate. In workplace B, one of the two engineered nanoparticle manufacturing workplaces, the PM 2.5 /TPM ratio was >0.7. This is possibly because the sampling point was located in front of a contained facility, and the door was open for bagging after nanopowder production was complete. A study by Tsai et al. (2011) suggested that when conducting exposure assessments at nanopowder-related workplaces, the mass distributions of both submicron (<1 µm) particles and respirable particles should be considered. Particle concentrations measured at the welding workplaces (workplaces C and D) were higher than at the engineered nanoparticle manufacturing workplaces. PM 2.5 generation ratios were above 90%, and nanoparticle number concentrations were very high. In this study, we found that particles generated during the welding process were almost all <2.5 µm. A study by Elihn and Berg (2009) also found that the percentage of submicron particles was high at high-temperature processes, but

42 that the number of nanoparticles was not higher than at other processes. Therefore, this suggests that PM 2.5 sampling at the welding process is appropriate for sampling the greatest concentration of generated particles. The size distributions measured by SMPS exhibited bimodal distributions at the unintended nanoparticle emitting workplaces. In most studies bimodal distributions were reported at the welding process; a bimodal distribution can be caused by the coalescence of nuclei (Vishnyakov et al., 2013). In addition, the size distributions differed between workplaces C and D. Particle sizes may have been affected by different welding methods and materials (Elihn and Berg, 2009). The metal contents at the unintended nanoparticle emitting workplaces were higher than those at the engineered nanoparticle manufacturing workplaces. Metal contents had lower concentrations compared to the total mass concentrations. In a study conducted by Yoon et al. (2009), metal contents, including Na and K, were analyzed, however we did not include Na and K in this study. Different welding materials can affect the chemical components and concentrations of particles (Yoon et al., 2009). In addition to processes that occur at the workplace, nanoparticles may also be generated by diesel engines (Vincent et al., 2000). Diesel-fueled forklifts were used in some workplaces for carrying products, and these machines can have a great effect on measurements. It is important to identify nanoparticle emission sources in the workplace. Therefore, Ham et al. (2012) suggest that researchers record activities in TAD

43 Other studies that have compared gravimetric and DustTrak PM 2.5 measurements have shown that the two metrics are well correlated, and that DustTrak concentrations were overestimated by a factor of 2 or more when compared to gravimetric PM 2.5 concentrations (Kim et al., 2004; Zhu et al., 2011). However, total dust (i.e., TPM) was measured in this study. There were a variety of differences between DustTrak concentrations and gravimetric sampling concentrations. It is estimated that the particle size range of the DustTrak is 1 15 µm, but the size range of the gravimetric sampling method is unlimited. It was difficult to collect sufficient particles when using time-interval sampling methods. Other studies have reported that short-term gravimetric sampling is difficult, and that the amount particulates in the short term would likely be below the LOD for gravimetric sampling (Kim et al., 2004). The minimum sampling time in low-concentration workplaces was over 2 hours in this study. Therefore, to collect enough mass using gravimetric sampling methods, long-term sampling is generally recommended. When comparing the time-integrated sampling method and direct-reading instrument sampling methods, the major advantage of direct-reading sampling was the ability to identify concentration variations. Concentration variation cannot be calculated with full shift gravimetric sampling because the sample size is one, however, with direct reading instruments it is possible to identify variation. When comparing the CV of 1 min intervals of direct-reading instrument data to that of time-interval gravimetric sampling, there were no significant differences in variations of sample size, analysis error, or characteristics of workplaces. However,

44 CV of 1 min interval in direct reading instruments was higher than CV of TIS interval in those. This result shows that, unlike gravimetric sampling methods, direct-reading instrument methods are able to identify concentration variations. Statistical results showed no differences between DustTtrak and gravimetric concentrations. All Spearman s rank test results indicated strong associations between gravimetric sampling and direct-reading sampling. But there are not significantly characteristics between Spearman s rank test and linear regression result. This result is considered small sample size. Table 6 describes the characteristics of online and offline monitoring according to workplace type. This study had several limitations. First, the workplaces we examined were not representative of all workplaces. Although various processes such as grinding, smelting, welding, and laser cutting can take place at unintended nanoparticle emitting workplaces, the processes we examined were primarily welding processes. Second, we were unable to perform a statistical comparison between the two types of nanoparticle generating workplaces because the engineered nanoparticle manufacturing workplaces had a small sample size. Third, particle concentrations in the workplace can be affected by season, ventilation, and working process. However, in this study we did not consider all of these factors

45 Table 6. Characteristics of online and offline monitoring results according to workplace type Online monitoring Offline monitoring Engineered nanoparticle manufacturing workplace As particles of various sizes are generated, direct-reading instruments such as SMPS, AeroTrak, or DustTrak are required to determine the particle size distribution As time-interval sampling is not suitable at workplaces with low particle concentrations, full shift gravimetric sampling is recommend for identifying mass concentrations Unintended nanoparticle emitting workplace To determine peak concentrations, sampling should be conducted at the welding process As most particles generated during the welding process are <2.5 µm, gravimetric sampling should be conducted

46 5 Conclusions In this study, concentrations measured at unintended nanoparticle emitting workplaces were higher than those at engineered nanoparticle manufacturing workplaces, and nanoparticle number concentrations were also high. PM 2.5 concentrations made up >90% of the total mass concentrations measured at welding workplaces, and most particles were <2.5 μm. Therefore, gravimetric methods should be used for sampling mass concentrations. However, mass concentrations measured at engineered nanoparticle manufacturing workplaces were relatively low, but the nanoparticle generation ratios were high. Therefore, both direct-reading instrument methods and gravimetric sampling methods should be used together at engineered nanoparticle manufacturing workplaces. There are many limitations in assessing short-term exposure and identifying variations in concentration. Although direct-reading instruments have limited accuracy and reliability, these instruments may be necessary to identify variations in concentration. Therefore, when conducting exposure assessments of workplaces we suggest the use of both direct-reading instrument sampling methods and gravimetric sampling methods together

47 6 References Brouwer, D. (2010). Exposure to manufactured nanoparticles in different workplaces. Toxicology. 269: Brouwer, D., Berges, M., Virji, M. A., Fransman, W., Bello, D., Hodson, L., Gabriel, S. and Tielemans, E. (2012). Harmonization of measurement strategies for exposure to manufactured nano-objects; report of a workshop. Ann Occup. Hyg. 56: 1-9. Brouwer, D. H., Gijsbers, J. H. J. and Lurvink, M. W. M. (2004). Personal exposure to ultrafine particles in the workplace: exploring sampling techniques and strategies. Ann. Occup. Hyg. 48: Conti, J. A., Killpack, K., Gerritzen, G., Huang, L., Mircheva, M., Delmas, M., Harthorn, B. H., Appelbaum, R. P. and Holden, P. A. (2008). Health and Safety Practices in the Nanomaterials Workplace: Results from an International Survey. Environ. Sci. Technol. 42: Elihn, K. and Berg, P. (2009). Ultrafine particle characteristics in seven industrial plants. Ann Occup. Hyg. 53: Ham, S., Yoon, C., Lee, E., Lee, K., Park, D., Chung, E., Kim, P. and Lee, B. (2012). Task-based exposure assessment of nanoparticles in the workplace. Journal of Nanoparticle Research. 14:

48 Heitbrink, W. A., Evans, D. E., Ku, B. K., Maynard, A. D., Slavin, T. J. and Peters, T. M. (2009). Relationships among particle number, surface area, and respirable mass concentrations in automotive engine manufacturing. J. Occup. Environ. Hyg. 6: Kim, J. Y., Magari, S. R., Herrick, R. F., Smith, T. J. and Christiani, D. C. (2004). Comparison of fine particle measurements from a direct-reading instrument and a gravimetric sampling method. J. Occup. Environ. Hyg. 1: Maynard, A. D. and Zimmer, A. T. (2002). Evaluation of Grinding Aerosols in Terms of Alveolar Dose: the Significance of Using Mass, Surface Area and Number Metrics. Ann. occup. Hyg. 46: Morawska, L., He, C., Hitchins, J., Mengersen, K. and Gilbert, D. (2003). Characteristics of particle number and mass concentrations in residential houses in Brisbane, Australia. Atmospheric Environment. 37: O'Shaughnessy, P. T. (2013). Occupational health risk to nanoparticulate exposure. Environmental Science: Processes & Impacts. 15: Ostiguy, C., Roberge, B., Woods, C. and Soucy, B. (2010). Current Knowledge about Occupational Health and Safety Risks and Prevention Measures, Vol. 2, IRSST Press, Montreal Paik, N.W., Park, D. W., Yoon, C. S. and Shin, Y. C. (2011). Workplace measurement and evaluation, Shinkwang books, Seoul

49 Spurny, K. R. (1998). Methods of Aerosol Measurement before the 1960s. Aerosol Science and Technology. 29: Tsai, C. J., Huang, C. Y., Chen, S. C., Ho, C. E., Huang, C. H., Chen, C. W., Chang, C. P., Tsai, S. J. and Ellenbecker, M. J. (2011). Exposure assessment of nano-sized and respirable particles at different workplaces. Journal of Nanoparticle Research. 13: Vincent, J. H. and Clement, C. F. (2000). Ultrafine particles in workplace atmospheres. The Royal Society. 358: Vishnyakov, V. I., Kiro, S. A., Ennan, A. A. (2014). Bimodal size distribution of primary particles in the plasma of welding fume: Coalescence of nuclei. Journal of Aerosol Science. 67: Yoon, C. S., Paik, N. W., Kim, J. H. and Chae, H. B. (2009). Total and Soluble Metal Contents in Flux-Cored Arc Welding Fumes. Aerosol Science and Technology. 43: Zhu, Y., Smith, T. J., Davis, M. E., Levy, J. I., Herrick, R., Jiang, H. (2011). Comparing gravimetric and real-time sampling of PM 2.5 concentrations inside truck cabins. J. Occup. Environ. Hyg. 8: Zimmer, A. T., Baron, P. A. and Biswas, P. (2002). The influence of operating parameters on number-weighted aerosol size distribution generated from a gas metal arc welding process. Journal of Aerosol Science. 33:

50 국문초록 작업장의나노입자상물질에대한 시간누적시료채취와실시간시료채취방법비교 김선주 서울대학교보건대학원 환경보건학과산업보건전공 지도교수윤충식 연구배경 : 나노입자는다양한분야의사업장및공정에서의도적, 비의도적으로발생한다. 기존의입자상물질의측정방법은중량법을이용하여평가하였지만, 나노입자의경우독특한특성때문에실시간기기를이용하여수농도혹은표면적농도로평가한다. 따라서기존의측정방법인시간누적시료채취와실시간기기를이용한시료채취방법을비교한연구가필요하다. 시간누적측정법인중량법은작업시간동안농도변이를알수없으며, 단시간측정이어렵다. 따라서본연구의목적은나노입자가발생하는사업장에서의입자상물질을평가하는전통적인방법과실시간기기를비교하고, 일반적인입자평가법인시간가중평균치방법과시간간격측정방법을비교하여작업장별적절한측정법을제안하는것이다

51 연구방법 : 측정은실시간기기인 SMPS, DustTrak, AeroTrak 및필터를사용하였다. 필터는실시간기기와동시에측정하였으며, 시간별비교를위하여전체작업시간동안측정및 1시간혹은일정시간간격마다교체하였다. 채취후분석은중량분석, 중금속분석, 전자현미경분석및시간별로측정한데이터비교를위하여통계분석을실시하였다. 결과 : 비의도적나노입자발생사업장이의도적나노입자발생사업장에비하여전체적으로많은입자가발생되었다. 농도의변이는시간간격을짧게측정할수록커졌다. 시간별로측정한필터와실시간기기는전체시간을측정한데이터가높은상관관계를보였다. 결론 : 용접사업장에는 PM 2.5 의이하의입자비율이높아필터를이용한중량법으로측정하여관찰하고, 의도적나노입자발생사업장에서는다양한입자크기분포를보이기때문에실시간기기와함께사용하는것이좋다. 그러나가장좋은방법은각각의장단점을보완하기위하여시간누적시료채취법과실시간기기를병행하여측정하는것을제안한다. 주요어 : 의도적나노입자, 비의도적나노입자, 실시간기기, 시간누적시료채취, 시료채취방법 학번 :